Performance inv & solve in Julia & R

[DNF]

@PharmCat

function f1(A,B)
    A\B
end
function f2(A,B)
    inv(A)*B
end
#=
ulia> @benchmark f1(A,B)
BenchmarkTools.Trial:
  memory estimate:  560 bytes
  allocs estimate:  4
  --------------
  minimum time:     19.501 μs (0.00% GC)
[...]
julia> @benchmark f2(A,B)
BenchmarkTools.Trial:
  memory estimate:  3.13 KiB
  allocs estimate:  11
  --------------
  minimum time:     1.120 μs (0.00% GC)

What version of Julia are you using? You have almost 20x performance difference here, between f1 and f2, that doesn’t seem right. I get only a 30% difference.

Also, when benchmarking, make sure to interpolate the variables in the benchmarking expressions:

@benchmark f1($A, $B)
@benchmark f2($A, $B)

Finally, if you are consistently using small matrices and vectors, where you know the sizes beforehand, you should definitely look at the StaticArray package, like @WschW suggests. You will probably get dramatic speedups that way, at least 10x I believe.